Home Articles The possibilities of identifying landuse categories on IRS data

The possibilities of identifying landuse categories on IRS data

ACRS 1996

Agriculture / Soil

The Possibilities of Identifying Landuse
Categories On IRS Data

U. Samarasena
Remote Sensing Technician,
Survey Department, Sri Lanka


Since 1980, the Remote Sensing of the Survey Department preparing Landuse maps using Aerial Photographs together with Satellite imagery’s by visual interpretation techniques. For effective monitoring of natural resources and updating of thematic maps Satellite Remote Sensing produces easy and efficient means, which is applies all over the World.

The high resolution satellite data (mainly SPOT and IRS) provides much information for identification of landuse categories. The LISS-II(Linear Image Self Scanner) with its improved Spatial and Spectral resolution enhanced these capability.

The major objective of the study describes herein is to use IRS satellite data with resolution 36m to identify the landuse patterns. The paper will discuss the methodology, results accuracy and the limitations.


The tropical island of Sri Lanka lies between 6 and 10 degree northern latitude to the southeast of the southern mosttip of the Indian subcontinent with an area of approximately 65600 square Km. The island consist of central highlands rising more than 2000m surrounded by extensive lowlands.

Due to the tropical climate and other physical parameters of Sri lanka natural climax vegetation on well draining land would be closed tropical forest and some of the forested area have been cleared for on going (or previous) developing projects. (e.g: Mahaweli Development work) That’s why the Landuse has been changing very rapidly.

On of the functions of the Sri Lanka Department (Center for Remote Sensing) is to provide landuse information of land development of the country. Such information not only covers agricultural land utilization, but also other non agricultural landuse activities such as Urban, Hydrology, Forested areas etc.

The application of Remote Sensing Technology in Center for Remote Sensing is mainly done by visual interpretation of black & white panchromatic aerial photographs. Conventional aerial photographs technique is time consuming and expensive as compared to satellite Remote Sensing and some of the information generated might have lost their value by the time the survey has been completed. That’s why a attempt is being made to use Computer Technologies with satellite data to achieve accurate and timely landuse information for various land development planning.

Objective of the Study
The Center for Remote Sensing has been producing landuse maps by interpreting Aerial Photographs and compiled on to satellite Imageries. This gives an accurate information for planning purposes. The objective of the study to assess the suitability and capability of IRS (LISS -II the sensor) data for landuse mapping and updating.

Materials and Method

The study was carried out based on 4 bands (3 band colour composite) of IRS clouds free data with the resolution of 36m acquired in March 1922 and the data format is BSQ. This data encloses the study areas in the south east part of Sri Lanka within the Uva Province. Digital data analysis was performed using PCI system at the Center for Remote Sensing of the Survey Department. The data was first Geometrically corrected to its real world position so, it could be correlated with the available references on the existing landuse maps which were prepared based on aerial photographs. Existing landuse maps and other related ground information as well as image enhancement techniques were used to assist in selection and delineation of training sites. The image then classified using part Maximum Likelihood Classification Algorithm.

Result and Discussion

The present Landaus classification legend of Sari lank based on 1:20000/1:50000 Aerial Photographs consist of 19 landaus classes which were grouped into 7 main categories. (see Appendix 1) The use of aerial photographs at this scale permit to mapping of details up to a unit size of 0.6 hectares (1:20000 scale)

However the use of 36m resolution IRS data only allows for a broad classification of the landuse patterns in the study area. As a result of landcovers were able to the classified and mapped.

Major features such as area under agricultural (sugarcane, paddy etc) Forested are and water bodies were easily identifiable on IRS data. Conversely more detailed (complexity areas) landuse parcels we not possible to identify due to the nature of agricultural land utilization in the area which is predominantly characterized by small sizes and irregular cropping patterns (These small areas consist of mixed cropping). However there are some categories such as sugarcane, irrigated paddy, water bodies and forested areas in large contiguous block were found to be mappable. Linear features such as roads, rivers and power lines also be possible to identify on the IRS data.

Some difficulties were encounter in the identification of paddy areas found within huge sugarcane growing areas. The spectral reflectance from these two crops almost similar during that particular time of data captured. This can be eliminated by using multitemporal data. Harvested sugarcane areas also be possible to identify by comparing with available landaus maps or ground truth.

Conclusions and Recommendations

Synoptic view of satellite data can be used for large areas of landuse categories such as paddy, Forest, Sugarcane etc. It is not possible to identify complex (or small) landuse pattern mostly with mixed cultivation. There will be a possibility to have high resolution satellite data in future with availability of new satellite observations. Even though it may be difficult identify and differentiate of very small landaus patterns. Hence, the algorithms that we used need to be improved or find better algorithms for such classification.

But this technique is more economical and less consuming as that’s done by the computer processing of satellite data. To get down satellite data from another country also takes time from the time of data captured as we don’t have receiving station in Sri Lanka. This a disadvantage for yield forecasing or estimating.

Still the aerial photographs are most suitable for landaus mapping, unless the resolution of satellite data should be around 5m of the future Remote Sensing Satellite System.


  • Final Report-Sri Lanka/Swiss Remote Sensing Project

  • Darus Ahmd: Application of Landsat MSS data in Landuse Mapping (ACRS 1989)